Mark Kim

Anthropic in the lens of Lilly

Pfizer is valued at $163B. Novo Nordisk at $142B.

Eli Lilly is worth a whopping $1.36T.

Eli Lilly spent $9.5B to date in R&D as of Q3 2026, but cash spent related to acquisitions, investments, and development milestones stood at $19B, equating to about $29B YTD or $40B for the FY.1

Anthropic spent $35B through Q3 on total operating expenses across 2 Fable-class and 5 Opus-class releases at an average of $5B per model (this is bad math since COGS of inference is included, but we can assume that inference is in turn used for development of the model, similar to running simulations for Teslas).

$33B has been generated by Anthropic in the same period.

Developing a new drug is estimated to require $3B. About 10% of Phase 1 trial drugs reach the market while 1% of drugs that make it to human trials make a positive financial return.

In this lens, Anthropic's valuation of $1.4T doesn't look too bad. Anthropic's models can be the engine of discovery that pumps out patentable research. The cost can also be amortized across different customers.

Eli Lilly likely has a higher hit rate than 1%, and Opus/Fable may not fully deliver something as valuable for a single generation of model. But Anthropic's model can be used by many companies, including Anthropic itself. What's more important is parlaying this technology development into something more. That's where the real opportunity and risk lies.

The UK developed titanium manufacturing capability whose distillation method was discovered in Luxembourg. France developed the transistor at a similar time as the US. Yet the European Concorde used an aluminum body. The compute industry didn't take off like the US in Europe.

Anthropic has shown that it can sell its technology, but transforming it into a higher-value product remains to be seen.

  1. Unusual year in that in FY 2025 Eli Lilly spent $19B.↩